Automatic and Effortful Encoding of Context

John F. Kihlstrom and William C. Heindel

University of Wisconsin

Research on the organic amnesic syndrome and other pathologies of memory has generated several lines of investigation on normal memory processes.  For example, early studies of amnesia suggested that patients were unable to retain any new information in long-term memory.  However, it is now clear  that amnesics can acquire an retain a large number of cognitive and motor skills -- although they may not be aware that they possess this knowledge.  Furthermore, amnesics show enhnced perceptual identification of targets drawn from a wordlist presented earlier, even though they cannot remember having studied the items.  Thus, amnesic patients appear to have most difficulty on tasks that require conscious awareness of the episodic context in which past events occurred.  Findings such as these raise questions concerning the manner in which the contextual features of episodes are encoded and retrieved in normal memory, and the relations between those forms of memory that require awareness of past experiences and those that do not.

In an influential paper, Hasher and Zacks (1979) proposed that information concerning the spatial and temporal context in which events occur is encoded automatically and effortlessly.  Evidence favoring this proposition came from findings that the respecification of spatial and temporal context is equivalent under intentional and incidental learning conditions.  On the other hand, Jacoby (1980) reported that respecification performance varied as a function of orienting task in a levels-of-processing procedure.  Since there is general agreement that automatic processes should be invariant across these sorts of study conditions, the Jacoby results appear to contradict those of Hasher and Zacks.  One purpose of our study was to investigate the encoding of spatial context information, comparing several different operational definitions of automaticity.

In our experiment, subjects were presented with 48 pairs of words, cues followed by targets, 12 of which appeared in each quadrant of a computer screen.  A total of 60 subjects were randomly assigned to one of three study conditions:

Hasher and Zacks compared the first two intentional learning conditions, but they did not include a truly incidental condition for comparison.

In both intentional learning conditions, the number and spacing of item-presentations was varied: 

The incidental learning condition incorporated a levels of processing manipulation:

Within each study condition, the items were roughly matched for imagery-value, meaningfulness, familiarity, and other attributes.  Half the items in each list were high, and half low, in word frequency.

Following the study phase, half the subjects in each study condition received an old-new recognition test consisting of the 48 targets plus an equivalent number of lures.  The remaining subjects received a perceptual identification test, in which targets and lures were presented for 33 msec, and the subjects were asked to identify the words.  Regardless of the type of test, after the presentation of each target the subjects were asked to indicate the quadrant in which the item had been presented.  For both tests, the computer presented the items and recorded responses and reaction times.

The recognition tests bear directly on the question of whether spatial context is encoded automatically or effortfully.  

The top panel of Table 1 presents the proportion of target items correctly recognized in the two intentional learning conditions combined.  There were significant (p.< .05) main effects for type of presentation (once, adjacent, or spaced) and for word frequency (high or low), but no difference between Item + Context and Item Only study conditions.

The bottom panel presents corresponding results for context recognition.  Again, there were main effects for presentation and frequency, but no effect of study condition and no interaction.  These results are in line with both Hasher and Zacks' and Jacoby's definitions of automaticity, in that respecification of context was unaffected by whether the subjects were instructed to remember the context in which the target occurred.  Taken alone, these results would indicate spatial context is,  indeed, encoded automatically.

However, this conclusion must be qualified by certain other findings.  Table 2 presents the recognition results for the incidental learning condition.  There was a main effect for frequency on target recognition (Table 2A), but not on context recognition (Table 2B).  More to the point, however, there was a main effect for orienting task.  With our very difficult physical task, there was no difference in recognition between it and the rhyme condition.  However, both target and context recognition were better in the semantic condition than in the other two.  Recall that according to both Hasher and Zacks' and Jacoby's definitions of automaticity, respecification of context should be invariant over encoding conditions.  These results, then, conflict with the earlier ones by indicating that spatial context is not encoded automatically after all.

Some resolution of this apparent discrepancy is provided by Table 3, which presents the overall recognition results for the three learning conditions, ignoring the various manipulations within each condition.  Both target and context recognition yielded significant effects for study condition, with recognition poorest following incidental learning.  Given the two definitions of automaticity, the finding of an effect of learning conditions on context recognition means, again, that context is not encoded automatically.  

To be sure, there is no difference between the two intentional learning conditions.  This may suggest that context is encoded routinely when subjects intend to recognize items, but this is not the same thing as an automatic encoding of context that occurs irrespective of task demands.  Thus, in the final analysis, spatial context appears to be encoded effortfully.  To the extent that the memory deficits shown by amnesics reflect a deficiency in encoding contextual information, the locus of their psychological deficit lies in those processes that require cognitive effort.  Truly automatic processes, then, should be spared in amnesic patients.

Among those memory tasks on which amnesics are unimpaired are those that do not require the respecification of the context in which events occurred.  For example, when patients are asked to identify tachistoscopically presented words, their performance is enhanced for words studied previously, even though they do not recognize these words as old.  This differential deficit suggests that there are two forms of memory, one that requires awareness of past experience and one that does not. Converging evidence from normal subjects was provided by Jacoby and Dallas (1981), who found that the number and spacing of repetitions affected both recognition memory and perceptual identification, but that level of processing affected only recognition.  They further concluded that there were two bases for recognition memory: perceptual fluency due to enhanced perceptual identification of an item, and respecification of an item's context.

We performed the same analyses on our data.  Recall that target recognition was affected by both presentation and levels of processing.  However, as Table 4 shows, perceptual identification showed an effect of number and spacing of repetitions (Table 4A), but not of level of processing (Table 4B).  Nor was there any difference in perceptual identification between the two intentional learning conditions.  This confirms the differential effects found by Jacoby and Dallas.  However, we took a closer look at this matter, by dividing the subjects' performance on both tasks into their respecification and fluency components.

Here is how the fluency and respecification components were separated.  

p(Respec_R) = p(Recog + Context)

p(Respec_I) = p(Ident + Context)

p(Fluen_R) = p(Recog - Context) = Recog - p(Respec_R)

p(Fluen_I) = p(Ident - Context) = Recog -  p(Respec_I)

Figure 1 shows the relative contributions of each component to recognition memory.  The top of each bar shows the proportion of items correctly recognized.  The bottom portion shows the magnitude of the fluency component, as we have defined it, while the top portion shows the magnitude of the respecification component.  

Figure 2 shows the corresponding analysis for perceptual identification.  

These findings have now been replicated in an experiment in which all subjects received both perceptual identification and recognition memory tests, in counterbalanced order.  We take these results as support for the idea that there are two bases of recognition memory -- onerepresented by a feeling of familiarity, the other by the respecification of episodic context.  Both bases are represented in the usual item-recognition test: if a subject cannot reconstruct the context in which an item has occurred, its prior occurrence can be inferred from a feeling of familiarity (Kihlstrom, 1984).  By the same token, it sometimes has been claimed that perceptual identification is a pure measure of fluency, because it is not affected by the sorts of variables that determine recognition memory.  It turns out, however, that perceptual identification is not a pure measure of fluency.  Both fluency and respecification contribute to perceptual identification.  The difference between identification and recognition lies in the relative importance of the two components.

Figure 3 makes the situation somewhat clearer.  In this figure, data for the high and low frequency words has been combined; again, the bottom portions of the bars represent fluency, while the top portions represent respecification.  As can be seen, the fluency component is more important in identification than in recognition.  Conversely, the respecification component is less important than in recognition.  

In perceptual identification, the subject is provided only with a degraded token of the word.  As a consequence, the subjects need fluency in roder to suggest possible decodings of the stimulus.  Subsequently, they can use respecification to help choose among the candidates generated by the fluency process.  Without fluency, however, respecification cannot be of help in perceptual identification.  

And, of course, respecification cannot help much in truly incidental learning situations, where the contextual information is unlikely to be encoded in the first place.  It may be of more help in intentional learning situations, where contextual information is available for use.  In recognition memory, the word is explicitly provided to the subject by the test.  For this reason, there is no perceptual benefit to be derived from fluency.  However, fluency is available as a heuristic for recognition, in case the subject is unable or unwilling to respecify the context in which the item occurred.

So where does this leave us?  There do indeed appear to be two forms of memory.  The one form, represented by the respecification component in any given memory task, requires the subject's conscious awareness of some prior episode.  the other form, represented by the fluency component, might be called preconscious.  We suspect that only conscious memory is affected by elaborative factors such as intentionality, level of processing, and repetition.  Amnesic patients appear to have a deficit in their conscious memory system, but may still be able to learn skills, acquire new factual information, and utilize this procedural and semantic knowledge through preconscious processes.  But even for normal subjects, it seems important to decompose memory performance into that component which reflects the subject's awareness of a prior episode, and that which does not.

Author Notes

Paper presented at the 25th annual meeting of the Psychonomic Society, San Antonio, Texas, November 1984.  This research was supported in part by a Biomedical Research Support Grant from the University of Wisconsin, and by Grant #MH-35856 from the National Institute of Mental Health, United States Public Health Service.  We thank Jeanne M. Sumi Albright, Arthur M. Glenberg, Irene P. Hoyt, Larry L. Jacoby, and Patricia A. Register for their helpful comments.  William C. Heindel is now at the Department of Neuroscience, University of California, San Diego.

References

Hasher, L., & Zacks, R.T.  (1979).  Automatic and effortful processes in memory.  Journal of Experimental Psychology: General, 108, 356-388.

Jacoby, L.L.  (1980).  Knowing and remembering: Some parallels in the behavior of Korsakoff patients and normals.  In L.S. Cermak (Ed.), Human memory and amnesia.  Hillsdale, N.J.: Erlbaum.

Jacoby, L.L., & Dallas, M.  (1981).  On the relationship between autobiographical memory and perceptual learning.  Journal of Experimental Psychology: General, 110, 306-340.

Kihlstrom, J.F.  (1984).  Posthypnotic amnesia and the dissociation of memory.  In G.H. Bower (Ed.), The psychology of learning and motivation (vol. 19).  New York: Academic Press.

This page last revised 04/08/10 02:58:48 PM.